Every hosting company has a status page. Green dots. Occasionally a yellow dot with "investigating" that turns green again 40 minutes later. You never find out what happened.
We got tired of that. So we built something different — a dashboard where each service tells you how it's actually feeling.
What is flame-emo?
Flame-emo (the name is short for emotional) is an internal operator dashboard that reads our watchman monitoring logs and derives a mood for each service: website, mail, DNS, panels, SSH. It shows up as a grid of cards — each one an emoji, a name, a label, and how long the service has been in that state.
The moods aren't arbitrary. They're computed from check history:
- 😊 All good — every recent check passed
- 😐 Mostly fine — back up, had a hiccup or two
- 😬 Recovered — up now but the failure rate was notable
- 🤔 Just stumbled — failed a check in the last few minutes, could be a blip
- 😟 Having issues — been failing for a while, this isn't a blip
- 😔 Not great — persistent failure, something upstream is broken
- 😱 Really struggling — long-running failure, please look at me
The last one being the most honest status message in the history of infrastructure monitoring.
You can talk to them
The part that actually surprised us in testing: you can open any service card and ask it a question. Type "what's going on?" and it responds — in first person, drawing on its own recent check history and the last 60 lines of its log file.
It sounds gimmicky. It isn't. When SMTP has been failing for 20 minutes and you ask it what's up, it tells you the last known error, when it started, and whether it's seen this before. That's just the log data surfaced in a readable way — but phrasing it as first-person makes you actually read it instead of scrolling past.
The AI that powers the chat is instructed to stay operationally precise. No drama, no excessive emoji, no hedging. If nothing is wrong in the logs, it says so plainly.
Why bother?
Monitoring tools are usually built for the moment things go wrong. You get an alert, you dig into logs, you piece together what happened. flame-emo is built for the moment before that — the ambient state of the system on any given afternoon.
Glancing at a grid of 😊 cards is genuinely reassuring. One 😟 in a sea of 😊 is immediately obvious. You don't have to run a check or open a terminal. You just look at it.
It also keeps historical mood context. A service that's currently up but was failing for two hours earlier today shows as 😬 Recovered, not 😊 All good. That distinction matters.
It lives in the operator panel
Flame-emo is an internal tool — it sits behind the operator panel's auth gate at /admin/emo. It's not customer-facing. Customers see the public status page; operators see the mood dashboard.
That's the right split. The public status page should be calm and factual. The internal dashboard can afford to be a bit more honest about the emotional state of the infrastructure.
If you're running your own infrastructure and want something like this, the core idea is simple: parse your monitoring logs, count recent pass/fail ratios, apply thresholds, render emoji. The AI chat layer is optional — the mood grid works fine without it.
The name was a joke that became accurate. Servers don't have feelings. But if they did, ours would at least tell you about them.